Introduction
HER2, an overexpressed cell-surface oncoprotein that contributes to breast, gastric, and other cancers [
1], is a validated therapeutic target, as evidenced by clinical success of the monoclonal antibody (mAb) trastuzumab [
2‐
5]. Trastuzumab acts against HER2-positive tumors by multiple mechanisms, including receptor internalization, receptor 'shedding', direct anti-proliferative activity, antibody-dependent cell-mediated cytotoxicity (ADCC), and presentation of antigenic determinants of opsonized cells to antigen-presenting cells [
6]. The latter mechanisms depend upon the interaction of the Fc domain of trastuzumab with Fc-gamma receptors (FcγRs) expressed by immune effector populations, such as natural killer (NK) cells or mononuclear phagocytes [
7‐
10]. Polymorphic variants of certain activating FcγRs predict response duration to trastuzumab: patients homozygous for CD16A (FcγRIIIA) 158V allele or CD32A (FcγRIIA) 131H allele or both have longer progression-free survival than patients carrying the respective 158F or 131R alleles [
11], which bind the Fc domain of immunoglobulin G 1 (IgG1), the main class of therapeutic mAbs, such as trastuzumab, with lower affinity than their allelic counterparts.
FcγR polymorphism influences the clinical response to several IgG1 mAbs other than trastuzumab. While the relationship between CD16A polymorphism and benefit is controversial for cetuximab [
12‐
15], CD16-158V and CD32A-131H homozygosity appear to be associated with beneficial responses for rituximab and infliximab [
16‐
18]. Furthermore, for an agonistic anti-death receptor antibody with intrinsic anti-tumor activity that is potentiated by FcγR interactions, effector cells expressing the higher-binding CD16A and CD32A variants supported substantially greater proapoptoptic activity [
19]. CD16A-158V homozygotes represent 10% to 20% of the population worldwide, whereas CD32A-131H homozygotes represent approximately 25% of Caucasians or Africans and 50% to 60% of Asians [
20,
21]. Thus, FcγR genotypes most frequently associated with greater beneficial responses occur in a minority of the population. This provides a strong rationale for engineering the Fc domain of trastuzumab to better interact with low-binding alleles of activating FcγRs to expand, without regard to FcγR genotype, the benefit of treatment to patients.
MGAH22 is an Fc-engineered mAb designed for increased binding to both alleles of CD16A and preservation of the direct anti-proliferative activity of trastuzumab. Since trastuzumab activity is enhanced in mice genetically deficient for the inhibitory FcγR, CD32B (FcγRIIB) [
7], a negative regulator of activation of monocytes, macrophages, and dendritic cells [
22], the Fc domain of MGAH22 was also engineered for reduced CD32B binding. The optimized Fc domain confers enhanced ADCC activity against HER2-positive tumors, including low HER2 expressors, independently of the FcγR variant for the effector cells. MGAH22 is active
in vitro and
in vivo against a HER2-positive tumor line derived from a patient whose tumor progressed while on trastuzumab. Because changes in effector cell interactions could have safety implications, high-dose MGAH22 toxicology studies were conducted in cynomolgus monkeys, a relevant species, with no significant antibody-related safety findings.
The enhanced properties of MGAH22 suggest potential clinical utility in extending the benefit of anti-HER2 immunotherapy to patients independently of their CD16A allelic expression and to patients who, because of low HER2 expression levels, do not qualify for trastuzumab treatment as well as to patients whose tumors progress while on trastuzumab.
Materials and methods
Human tumor cell lines
Breast (MCF-7, ZR-75-1, and SKBR-3), gastric (N87), colon (HT-29), and bladder (SW780) lines were obtained from the American Type Culture Collection (Manassas, VA, USA), and JIMT-1 (breast) was obtained from DSMZ (Braunschweig, Germany). All cell lines were cultured in accordance with recommended specifications for fewer than 30 passages. The number of HER2-binding sites per cell and immunohistochemistry category (0 to 3+) were determined by flow cytometry (QuantiBRITE™ PE; BD Biosciences, San Jose, CA, USA) and HercepTest™ (Dako, Carpinteria, CA, USA): MCF-7 (14,000; 1+), HT-29 (25,000; 1+), SW780 (37,000; 1+), ZR-75-1 (52,000; 2+), JIMT-1 (79,000; 2+); N87 (270,000; 2+), and SKBR-3 (540,000; 3+).
Antibodies
ch4D5 was generated by fusing synthetic sequences encoding light- and heavy-chain variable domains of 4D5, the murine precursor of trastuzumab [
23], to human κ and IgG1 constant domains, respectively. RES120 was generated from ch4D5 by light-chain mutagenesis (N65S) to eliminate a consensus N-glycosylation site. MGAH22 was generated from RES120 by exchanging its Fc domain for MGFc0264 (L235V, F243L, R292P, Y300L, and P396L) [
24,
25]. ch4D5-N297Q, which contains an inactivated Fc domain, was derived from ch4D5 by mutating the heavy-chain N-glycosylation site.
Fc-gamma receptor binding
Binding of soluble forms of FcγRs (either monomeric extracellular domains or dimeric inactivated Fc-G2 fusions) to Fc domains was analyzed by surface plasmon resonance after capture of antibodies to immobilized HER2 [
25].
In vitro anti-proliferation activity
Tumor cells (2 × 104 per well) were incubated for 6 days with antibodies at 37°C, and proliferation/viability was detected by using the CellTiterGlo Luminescent Cell Viability Assay Kit (Promega Corporation, Madison, WI, USA).
Antibody-dependent cell-mediated cytotoxicity
Peripheral blood mononuclear cells (PBMCs) were isolated from healthy human donor blood (Ficoll-Paque™ Plus; GE Healthcare, Piscataway, NJ, USA). NK cells were purified (Untouched Human NK cell isolation kit; Dynal, Invitrogen Corporation, Carlsbad, CA, USA). Target cells (2 × 104 per well) were incubated with antibodies for 30 minutes at 37°C in RPMI-1640 (without phenol red), 10% fetal bovine serum, and 2 mM GlutaMax™ (Invitrogen Corporation) before adding effector cells at an effector/target ratio of 30:1 (PBMCs) or 3:1 (purified NK cells). Lactate dehydrogenase release (Promega Corporation) was measured after overnight incubation. Cytotoxicity (expressed as a percentage) = (experimental cell lysis-antibody-independent cell cytolysis)/(maximum target lysis-spontaneous target lysis) × 100. FcγR genotypes were determined by sequencing polymerase chain reaction-amplified DNA.
Treatment of xenograft tumors in mice
All mouse experiments were performed at our facility under protocols approved by the MacroGenics Institutional Animal Care and Use Committee. RAG2-/- Balb/c mice (WT FcγR mice) were obtained from Taconic (Rockville, MD, USA). mCD16-/- RAG2-/- Balb/c mice (mCD16 knockouts) and mCD16-/- hCD16A+ RAG2-/- mice (mCD16 knockouts transgenic for hCD16A-158F) were bred at MacroGenics. JIMT-1 cells (5 × 106 per mouse) in phosphate-buffered saline plus Matrigel were implanted subcutaneously and antibodies administered intraperitoneally weekly beginning at the time of tumor implantation or after tumors of approximately 200 mm3 had been allowed to form. Tumor sizes were monitored three times weekly by orthogonal measurements with electronic calipers. Statistical differences in tumor sizes were assessed by two-way analyses of variance and Bonferroni post-test analyses (GraphPad Prism 5.02; GraphPad Software, Inc., La Jolla, CA, USA).
Toxicology/toxicokinetics in non-human primates
Cynomolgus monkey experiments were conducted at Charles River Laboratories (Sparks, NV, USA) in accordance with Testing Facility Standard Operating Procedure, which adheres to the regulations outlined in the US Department of Agriculture Animal Welfare Act [
26] and the conditions specified in the
Guide for the Care and Use of Laboratory Animals [
27]. The study protocols were approved by the Testing Facility Institutional Animal Care and Use Committee. A single-dose study was conducted with 12 cynomolgus monkeys randomly assigned to two groups (three per sex per group) receiving MGAH22 or RES120 at 50 mg/kg by 60-minute intravenous infusion. All animals were euthanized at day 62 for necropsies. A repeat-dose study was conducted with 40 monkeys randomly assigned to four groups (five per sex per group) receiving vehicle or MGAH22 weekly for 6 weeks at 15, 50, or 150 mg/kg by 60-minute intravenous infusion. Twenty-four (three per sex per group) were euthanized on day 40, 4 days after the last dose, and 16 (two per sex per group) were followed for a recovery period of 56 days and euthanized on day 93 for necropsies.
Measurement of serum MGAH22 concentration by enzyme-linked immunosorbent assay
Goat anti-MGAH22-Fv antibody was used to capture MGAH22 (or RES120) from cynomolgus serum and this was followed by detection with biotinylated goat anti-MGAH22-Fv antibody plus streptavidin-alkaline phosphatase conjugate and 4-methylumbelliferyl phosphate as substrate. Product was measured by using a fluorescent microplate reader against a standard curve (four-parameter non-linear curve fitting). The minimum quantifiable concentration was 4.25 ng/mL.
Measurement of cytokine release
BD™ human Th1/Th2 cytometric bead array (CBA) and Human IL-5 Flex Set CBA were used to measure levels of IL-2, IL-4, IL-5, IL-6, IL-10, tumor necrosis factor-alpha (TNF-α), and interferon-gamma (IFN-γ) in serum collected at different times after intravenous administration of antibodies to cynomolgus monkeys. Statistical differences in relative changes from baseline were assessed by non-parametric van Elteren tests (an extension to the Wilcoxon rank-sum test) and Bonferroni adjustments for multiple comparisons (SAS 9.2; SAS Institute Inc., Cary, NC, USA). Cytokine levels were also measured in cell culture supernatants after incubation of human PBMCs with antibodies on plates left uncoated or coated with 1 μg/mL of HER2 antigen-recombinant human ErbB2/HER2 Fc chimera (R&D Systems, Inc., Minneapolis, MN, USA), enzymatically deglycosylated-for 16 to 20 hours at 37°C. Statistical differences in cytokine levels induced by the different antibodies were assessed by a Wilcoxon signed rank test (SAS 9.2).
Discussion
MGAH22 is a human/mouse chimeric IgG1 anti-HER2 antibody based on mouse clone 4D5, the precursor to trastuzumab. MGAH22 was engineered to maintain the antigen-binding properties of the original antibody while optimizing its interactions with human FcγRs, important mediators of antibody function in vivo. The engineered Fc domain of MGAH22 imparts increased affinity for both allelic variants of the low-affinity activating FcγR, CD16A, and decreased affinity for the inhibitory FcγR, CD32B. While maintaining the HER2-binding properties and direct anti-proliferative activity of trastuzumab against sensitive cell lines, these enhanced binding properties confer additional improvements in terms of enhanced anti-tumor activity against HER2-expressing tumor cell lines in vitro; the greatest improvement was observed in ADCC activity against the lower (1+ and 2+) HER2-expressing cell lines and/or with effector cells isolated from human donors homozygous or heterozygous for the low-binding allele (158F) of CD16A. In vivo, MGAH22 exhibited enhanced anti-tumor activity against a 2+ HER2-expressing cell line in mice genetically deficient for murine CD16 but transgenic for the human CD16A-158F variant.
The FcγR-binding profile of MGAH22 has many unique aspects. The increased binding affinity to CD16A compares well with that observed with afucosylated trastruzumab [
36] or other Fc-engineered mAbs [
37‐
39]. The improvement in MGAH22 binding to the 158F allele of human CD16A to levels exceeding those of the WT Fc domain for the 158V allele suggests that MGAH22 could provide benefit to patients of any CD16A genotype, but particularly homozygotes or heterozygotes carrying the 158F variant, who have poorer outcomes in response to trastuzumab treatment [
11]. The increased CD16A binding also resulted in increased effectiveness, particularly against tumor cells expressing low levels of HER2. This suggests that MGAH22 can induce productive synapse formation between tumor and effector cells with fewer antibody-target interactions on the tumor cell surface, presumably by recruiting more Fc receptors on the effector cells per unit of binding or increasing the length of time the receptor is engaged or both. Because the benefit of trastuzumab therapy accrues only to patients with tumors that overexpress HER2 at the 3+ level or exhibit gene amplification [
3,
5,
40], this finding suggests that MGAH22 may extend such advantages of anti-HER2 therapy to patients whose tumors express low HER2 levels and who are not thought to benefit from trastuzumab treatment.
CD16A is coexpressed with other FcγRs on mononuclear phagocytes but is the only FcγR expressed by NK cells. These cells are the major contributors to ADCC activity in PBMCs under standard
in vitro assay conditions, a notion supported by the observation that enhanced MGAH22-mediated ADCC was also observed with purified NK cells. NK cells have been implicated as important mediators of the anti-tumor activity of trastuzumab in patients with breast cancer. Trastuzumab treatment is associated with increased numbers of tumor-associated NK cells, and patients with responsive tumors tend to have larger numbers of tumor-infiltrating NK cells [
8,
9]. Patients with higher NK cell numbers exhibit higher levels of trastuzumab-mediated ADCC activity, which has been associated with increased tumor responsiveness [
9,
10]. This finding is consistent with the association between responsiveness to trastuzumab treatment and level of ADCC activity mediated by CD16-expressing cells and CD16A genotype [
11].
Another unique feature of the Fc domain of MGAH22 is its decreased binding to the CD32B inhibitory receptor. Fc domains exhibiting decreased fucosylation, by comparison, enhance binding only to CD16A [
39,
41], and other mutations reported to increase binding to activating receptors also demonstrate increased binding to CD32B but to different extents, depending on the mutations [
37,
38]. When co-engaged with an activating FcγR on mononuclear phagocyte effectors, CD32B confers an inhibitory signal that counters cell activation. Although no clinical data on an association between reduced CD32B binding and response to trastuzumab or other mAbs are available, non-clinical studies show the importance of Fc-mediated functions exerted by monocytes and macrophages
in vivo [
42]. Enhanced anti-tumor responses occur in mice genetically lacking CD32B [
7], and enhanced antigen delivery via immune complexes that bind both activating and inhibitory receptors occurs under conditions of CD32B blockade [
43‐
45]. These effects may contribute to the ability of immunotherapy to break tolerance in cancer and induce an adaptive immune response. Attempts at modeling the CD32B-dependent component of MGAH22 action in terms of effector cell function have been hampered by the ineffective tumor cytotoxic activity of mononuclear phagocytes
in vitro (data not shown) and lack of a suitable animal model. Nonetheless, a decline in binding to CD32B is expected to be beneficial by increasing the ratio of activating-to-inhibitory FcγR interactions.
In the selection of species for non-human primate toxicology studies, both antigen expression and Fc/FcγR interactions were considered. Tissue cross-reactivity studies with MGAH22 on human and cynomolgus tissue panels revealed similar antigen distributions, which were comparable to those observed with trastuzumab. Importantly, the binding profile of MGAH22 to cynomolgus monkey FcγRs generally supports the use of this species as a relevant toxicology model for the immune effector function of this antibody. Although the engineered Fc domain has increased binding to cynomolgus monkey CD32B, which differs from its decreased binding to human CD32B and may limit toxicity in monkeys, it has increased binding to the invariant monkey CD16A and CD32A receptors. Moreover, the binding affinities for these activating FcγRs of monkeys exceed those for the high-binding alleles of the human orthologs, a situation that may counteract the potential inhibitory effect of increased binding to monkey CD32B and be adequate for evaluating potential toxic effects due to FcγR engagement.
The MGAH22 Fc domain preserves FcRn binding, which favors an extended serum half-life [
46]. The terminal half-life of MGAH22 in cynomolgus was 7 to 9 days, approximately 20% shorter than that of RES120, which contains the WT Fc domain. A similar decline in half-life was observed when afucosylated trastuzumab, exhibiting increased binding to hCD16A, was compared with trastuzumab in hCD16A transgenic mice [
36]. Except for a slightly shorter half-life, the pharmacokinetic profile of MGAH22 in cynomolgus monkeys is comparable to that of other anti-HER2 mAbs [
34,
35]. Importantly, in the single- and repeat-dose toxicology studies, there were no significant antibody-related clinical observations or macro/microscopic findings. The modest dose-independent decrease in circulating NK cells was reminiscent of a similar observation in monkeys treated with an Fc domain-enhanced anti-CD19 mAb [
37]. Given its transient nature, the phenomenon likely results from margination of the NK cells. The Fc-engineered MGAH22 mAb was not unusually immunogenic in monkeys, but owing to the lack of predictive value of immunogenicity data in animals [
47], the potential incidence of immunogenicity in humans cannot be extrapolated.
Cytokine release could be exacerbated by increased binding to FcγRs. MGAH22 induced minimal levels of just IL-6 in cynomolgus monkeys and IL-6, TNF-α, and IFN-γ from human PBMCs
in vitro that were similar to those induced by RES120 or trastuzumab, suggesting that MGAH22 is unlikely to induce cytokines in patients to levels any higher than those induced by trastuzumab. A potential explanation is that cytokine release may relate more to CD32A than CD16A. CD32A expression by mononuclear phagocytes, but not NK cells, is consistent with the spectrum of observed cytokines, which did not include IL-2, an NK cell-derived cytokine. Binding of MGAH22 to the prevalent CD32A-131H allele is unchanged compared with WT Fc domains, whereas binding to the rarer CD32A-131R allele is decreased, a reflection of the high degree of homology between the extracellular domain and Fc-binding interface of this variant with CD32B (including the arginine at position 131, which is shared by CD32B). While CD32A polymorphism may contribute to outcomes in patients with trastuzumab-treated metastatic breast cancer, its role is less pronounced than that associated with CD16A polymorphism [
11]. Moreover, recent data suggest an association of the 131H allele of CD32A with the development of trastuzumab-related cardiotoxicity [
48]. Thus, the lack of enhanced binding to either of the CD32A alleles may be favorable to the safety profile of MGAH22.
Other novel HER2-directed agents are undergoing clinical development. A trastuzumab-drug conjugate, T-DM1, designed to deliver a cytotoxic molecule into HER2-overexpressing cells via receptor-mediated endocytosis [
49], has shown a significant advantage in advanced breast cancer, although its benefits appear to be restricted to patients with HER2 3+ or gene-amplified tumors [
50]. In this context, MGAH22 may have particular utility in patients with low HER2-expressing tumors. An afucosylated version of trastuzumab with increased anti-tumor effector function has also been described [
36]. However, MGAH22, by exhibiting diminished binding to the inhibitory FcγR, CD32B, differs from afucosylated trastuzumab, which exhibits a slight increase in binding to this inhibitory receptor. MGAH22, by diminishing interactions with this inhibitory FcγR, would be expected to exhibit additional favorable properties in the presence of mononuclear phagocytic effector cells and potentially further enhanced efficacy against low HER2-expressing tumors or tumors resistant to trastuzumab therapy.
Competing interests
All authors are or have been employed by MacroGenics, Inc., a privately held company, and have received MacroGenics stock options as a condition of employment.
Authors' contributions
SG, WZ, YY, HL, SB, LH, VC, and TZ conducted experiments and helped to analyze data. SJ, PAM, JS and SK helped to conceive and design experiments and analyze data. SJS helped to analyze data. EB and JLN helped to conceive and design experiments, analyze data, and write the paper. All authors read and approved the final manuscript.